)]}'
{
  "log": [
    {
      "commit": "9a80a7326b05448284c16df683f5ba0e148d4774",
      "tree": "aebf558ae2e05fc5fb3d0203809b2a7a18f6b80f",
      "parents": [
        "7a8515d72704f9fbb196b34bf3a9bc5d5dbdd137"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 06:26:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 20 06:57:17 2026 +0100"
      },
      "message": "[PERF] RawFieldParser: copy a run at once and test stop chars with bit masks\n\ncopyContent and copyUnquotedContent appended one char at a time to a builder\nof 16 chars and ran BitSet#get, isWhitespace and the \u0027(\u0027 / \u0027\"\u0027 tests per byte.\nThe run is now found first, with the delimiters, whitespace, \u0027(\u0027 and \u0027\"\u0027 as\ntwo longs for ASCII (BitSet#get remains for a non-ASCII delimiter), then\nappended at once. CR and LF are still deleted and a run holding a non-ASCII\nbyte is still decoded as UTF-8. The builders of parseValue, parseDomain and\nparseLocalPart are sized from the cursor.\n"
    },
    {
      "commit": "7a8515d72704f9fbb196b34bf3a9bc5d5dbdd137",
      "tree": "83f700001b0db5a4e4b4bb79bbe0dc6577f8154e",
      "parents": [
        "c9d4e103ff1feecfdd2df8546ef96d7531d04c12"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 06:26:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 20 06:57:17 2026 +0100"
      },
      "message": "[PERF] QuotedPrintableInputStream: write plain bytes and single spaces directly\n\nEvery plain byte went through transfer() and every space through the blanks\nbuffer. A plain byte (above space, not \u0027\u003d\u0027) with no blank held back and no\npending CR is now written as is, and so is a space followed by a plain byte,\nwhich cannot be a trailing blank. Character.isWhitespace on a byte is replaced\nby a table built from it. Trailing blanks, soft line breaks, long blank runs\nand read-size independence take the paths they took.\n"
    },
    {
      "commit": "c9d4e103ff1feecfdd2df8546ef96d7531d04c12",
      "tree": "01d8f18fa3bb62fa12d134afda543f18d3570394",
      "parents": [
        "fc01ffa74d1be6d637639ef06d06c33e91cd79e3"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 06:26:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 20 06:57:17 2026 +0100"
      },
      "message": "[PERF] Base64OutputStream: encode whole lines with the JDK array encoder\n\nThe wrapped JDK stream encoder encodes in Java and writes each base64 char with\nout.write(int): 70% of writeMessage. Blocks of whole lines (about 3 KB) are now\nencoded with Base64.Encoder#encode(byte[], byte[]), which is intrinsic, and\nwritten in one call. A partial line or a carry goes through a group encoder\nwriting to a buffer, flushed once per write call.\n\nThe output is byte for byte the same: line length rounded down to a multiple\nof 4, separator between lines, a trailing separator only when something was\nwritten. A test compares it to the JDK stream encoder for random sizes, write\nsplits, line lengths and separators.\n"
    },
    {
      "commit": "fc01ffa74d1be6d637639ef06d06c33e91cd79e3",
      "tree": "5bf83d35587d07c2ce2b3d2b76d58cfc2c0eaebe",
      "parents": [
        "06ca47fd6a4462799b1fc4bf1a98a655f385bf84"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 06:26:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 20 06:57:17 2026 +0100"
      },
      "message": "[PERF] DecoderUtil: feed the Q and B decoders a byte array and read them in bulk\n\nEach encoded word went through a TextInputStream (a CharsetEncoder and a 1 KB\nbuffer) to turn an ASCII String into bytes, and the decoder was then read one\nbyte at a time. The encoded text is now taken with String#getBytes(US_ASCII)\n(\u0027?\u0027 for a char out of US-ASCII, as the replacing encoder did), decoded by the\nsame QuotedPrintableInputStream / Base64InputStream, and read in bulk into an\narray of the encoded length. The \u0027_\u0027 of Q is replaced by \u003d20 on the bytes.\n"
    },
    {
      "commit": "06ca47fd6a4462799b1fc4bf1a98a655f385bf84",
      "tree": "b7e146b2a9b3e0d8caf2c627efb173795ffa8107",
      "parents": [
        "f011ca9e5ca7129f9842b59b7701801c50ffb54f"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 06:26:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 20 06:57:17 2026 +0100"
      },
      "message": "[PERF] MessageMatcher: read the searched content in blocks\n\nReader#read() decodes one char per call: on an InputStreamReader each char took\nthe stream lock twice, allocated a char[2] and a CharBuffer and ran the charset\ndecoder, over 60% of matchBody. The content is now read 4096 chars at a time;\nthe matcher state carries across blocks as it did across chars, and is still\nreset once per reader. The needle is kept in a char[] and the matchers are\nwalked by index.\n"
    },
    {
      "commit": "f011ca9e5ca7129f9842b59b7701801c50ffb54f",
      "tree": "b4739d3ce65e9768c1a5144305a35a132f336288",
      "parents": [
        "7b9bfed84a8267999fb53cf60e9864dffb06e6ca"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 18 22:37:34 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 05:19:52 2026 +0100"
      },
      "message": "[ENHANCEMENT] MimeUtil: draw boundaries and Message-IDs from a SecureRandom\n"
    },
    {
      "commit": "7b9bfed84a8267999fb53cf60e9864dffb06e6ca",
      "tree": "4bca8c96cb76a33159074d6210e69d32c65c38f2",
      "parents": [
        "082726140b4dc8f7f8bf036d58244e16495b8a9a"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 18 22:37:20 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 05:19:52 2026 +0100"
      },
      "message": "[PERF] FromLinePatterns: match a long line in linear time\n"
    },
    {
      "commit": "082726140b4dc8f7f8bf036d58244e16495b8a9a",
      "tree": "c88e90c723e2f094ea204d3a55c891dc16346a5d",
      "parents": [
        "dac6bbdd9bf7464e0d2c9101cc1317ac79e6ccd5"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 18 22:35:06 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 19 05:19:52 2026 +0100"
      },
      "message": "[PERF] MimeUtil#fold: look for a fold at the end of a whitespace run once per run\n"
    },
    {
      "commit": "dac6bbdd9bf7464e0d2c9101cc1317ac79e6ccd5",
      "tree": "11b267f3b2add90548de1beb1fd64b07744bfa76",
      "parents": [
        "c69a126d28f6f5bc0dac6b781aa1321b75635a45"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 17 09:37:09 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 17 10:16:35 2026 +0100"
      },
      "message": "[ENHANCEMENT] LenientDateTimeFieldTest should be more predicatble\n"
    },
    {
      "commit": "c69a126d28f6f5bc0dac6b781aa1321b75635a45",
      "tree": "ab9b97b6bef831293de1d8ee02ca2658183bbcb6",
      "parents": [
        "87de46594d00d32ed9395c10c3b03dd076f176ed"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 15 21:49:21 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] AddressListParser: an atom of one non-ASCII character is a DOTATOM\n\nSince the RFC 6532 support of MIME4J-331, UTF8NONASCII was declared as a\npublic token, before DOTATOM. On a single non-ASCII character followed by\na delimiter (\"@\", space, \"\u003c\", \",\"), both tokens match one character and\nJavaCC picks the first declared: UTF8NONASCII, which no production uses.\nSo the strict parser (DefaultAddressParser) rejected \"é@example.com\",\n\"a@é\" and \"Jean é \u003cjean@example.com\u003e\", while it accepted longer UTF-8\natoms such as \"müller\", lexed as the longer DOTATOM match.\n\nUTF8NONASCII is now private, as ALPHA, DIGIT and ATEXT: it is only a part\nof ATEXT, and a one-character atom is a DOTATOM whatever its characters.\n\nFixes #136\n"
    },
    {
      "commit": "87de46594d00d32ed9395c10c3b03dd076f176ed",
      "tree": "280cae20cc5e63000b594256430af6abb2771dab",
      "parents": [
        "c16049e0b879429a82b3c55481dcad8e19400eab"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 14:04:22 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] MessageMatcher: header fields are bounded, and maxTotalHeaderLen bounds a field when maxHeaderLen is off\n\nMessageMatcher, which James uses on untrusted mail for the Sieve body test\nand the in-memory BODY / TEXT search, parsed with setMaxHeaderLen(-1).\nMimeEntity#readRawField derives its line bound from maxHeaderLen, so a\nsingle field of N bytes, on one line or folded over many, was held whole\nin the line buffer and the field builder, each doubling, then decoded to a\nString, before maxTotalHeaderLen rejected it once the field was built.\n\nMessageMatcher now keeps body lines unbounded but bounds header fields at\nthe MimeConfig.PERMISSIVE header length (64 KB), and a field over it ends\nthe search without a match instead of failing the caller.\n\nIn core, when maxHeaderLen is turned off, maxTotalHeaderLen, which a\nsingle field cannot exceed either, bounds the header line read and the\ndefault field builder: the message wide limit stops the buffering instead\nof applying once the field is built.\n"
    },
    {
      "commit": "c16049e0b879429a82b3c55481dcad8e19400eab",
      "tree": "a690ada0d29a6504b409babe499a38c66d484852",
      "parents": [
        "6c6f886b78913d97d4946dafc6e31032cfc3b288"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 14:02:27 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] QuotedPrintableInputStream: decoded content is served from an offset, not shifted on every read\n\nBlanks are held back so that trailing blanks before a line break can be\ndropped. When a run of them ended with a visible byte, what did not fit in\nthe caller\u0027s buffer went to decodedBuf, and each later read served at most\nlen bytes from its front and then shifted all the rest with\nremove(0, chunk). Draining N bytes with reads of L bytes moved about\nN^2 / 2L bytes: with the 4 KB reads of ContentUtil#copy, some 12 GB of\nmemmove for a part of 10 MB of spaces followed by one letter, on each\ndecode of the message.\n\ndecodedBuf is now served from a read offset, cleared once drained, so the\ndrain is linear. It is only appended to once the caller\u0027s buffer is full,\nwhen it has already been drained: the decoded output does not change.\n"
    },
    {
      "commit": "6c6f886b78913d97d4946dafc6e31032cfc3b288",
      "tree": "8acf93e5fd0e34828a040ca73bffbf9724b47f01",
      "parents": [
        "335c95a8cd1f1d20756dd5b85beb9f54bcec2f4b"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 14:00:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] MimeEntity: a header line over maxHeaderLen is rejected whatever the read sizes\n\nSince 2242af35, readRawField threw MaxHeaderLengthLimitException before\nemitting the current field only when readLine cut the over-long line\nshort, that is when the read crossing the bound held no LF. When the LF\ncame in that same read, the line was read whole, the field before it was\nemitted, and the same exception came one event later from\nDefaultFieldBuilder#append. So the same message gave one T_FIELD more or\nless depending on how the source split its reads (network segments, IMAP\nliteral chunks), in the message header and in body part headers.\n\nreadRawField now rejects any line whose length, counted as\nDefaultFieldBuilder#append counts it (CRLF as one byte), is over\nmaxHeaderLen, whether it was cut short or ends with LF: such a line cannot\nbelong to an accepted field, and the field before it is never emitted.\n\nFixes #134\n"
    },
    {
      "commit": "335c95a8cd1f1d20756dd5b85beb9f54bcec2f4b",
      "tree": "064cf4a102374ef9a0c688afd42822a4096aef2c",
      "parents": [
        "dd60f6708d82fd3b500a4c4263941f72f4d68d5a"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 13:59:15 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] ContentTypeFieldImpl: a quoted parameter value starting with \"\u003d?\" is decoded as the lenient reader does\n\nSince the fix of #126, EncoderUtil#encodeHeaderParameter writes a\nContent-Type parameter value starting with \"\u003d?\" as one quoted B encoded\nword, which the lenient reader decodes. The strict ContentTypeFieldImpl\n(DefaultFieldParser, strict DefaultMessageBuilder, DefaultBodyDescriptorBuilder\nby default) kept the quoted-string as is: Fields.contentType(\"text/plain\",\nname\u003d\"\u003d?utf-8?q?x?\u003d\") read back as \"\u003d?utf-8?B?PT91dGYtOD9xP3g/PQ\u003d\u003d?\u003d\".\n\nA parameter value starting with \"\u003d?\" can only come from a quoted string, as\n\u0027\u003d\u0027 and \u0027?\u0027 are not token characters: the strict reader now decodes it as\nencoded words, as RawFieldParser and many mailers do (name\u003d\"\u003d?...?\u003d\"). The\nboundary, a byte exact delimiter, is still never decoded. Both readers then\nagree, on built fields and on received ones.\n\nFixes #133\n"
    },
    {
      "commit": "dd60f6708d82fd3b500a4c4263941f72f4d68d5a",
      "tree": "d976ebd237d417e98f45249dfbe35f1269ea94d5",
      "parents": [
        "cd29e0789a0522be127b0ac654dcc9cacadc47e7"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 13:58:06 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] DateTimeFieldLenientImpl: seconds after white space or a comment are read\n\nRFC 5322 obs-hour, obs-minute and obs-second allow CFWS around the colons\nof the time. On \"09:55: 06 -0600\" or \"09:55:(x)06 -0600\" the optional\nseconds section of RFC_5322 failed, every later section too, and the parse\nstopped at the second colon. zoneLeftUnread only rejects a rest starting\nwith a sign or a digit, so \": 06 -0600\" was ignored and the date read as\n09:55:00 GMT, hours away from the instant of the strict parser: a wrong\nsent date for IMAP SENTBEFORE/SENTSINCE, JMAP sentAt and sort order.\n\nWhite space and comments around a colon between two digits are now\nremoved before parsing, so the seconds and the zone are read as the strict\nparser reads them. A value whose parse still stops on a \u0027:\u0027 or a \u0027.\u0027 (a\nnested comment) gives no date rather than one without its seconds and\nzone.\n\nFixes #131\n"
    },
    {
      "commit": "cd29e0789a0522be127b0ac654dcc9cacadc47e7",
      "tree": "9ecc8cd54bcf156732150e5fe8478b943f613398",
      "parents": [
        "fd839dc0b685dd4801d3e488ba7c89d8e5df8fa3"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 13:56:33 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] LenientAddressParser: the quoted local part of a mailbox without domain is not decoded\n\nFor an address with no \"@\" nor \"\u003c\", createMailbox RFC 2047-decoded the\nwhole opening text, whose quotes parseValue had already removed: the\ncontent of a quoted string was decoded like bare text. The fix of #125\ncovered the local@domain and angle-addr forms only. So the domainless\nmailbox \"a \u003d?utf-8?q?x?\u003d\", which AddressFormatter writes as\n\"\\\"a \u003d?utf-8?q?x?\u003d\\\"\", was read back as \"a x\", and a quoted text decoding\nto nothing, such as \"\\\"\u003d?x???\u003d\\\"\", dropped the mailbox from the list.\n\nAs in the local@domain form, a domainless text holding a quoted string is\nread again as a local part, without decoding. Bare text keeps the decoding\nof MIME4J-253.\n\nFixes #130\n"
    },
    {
      "commit": "fd839dc0b685dd4801d3e488ba7c89d8e5df8fa3",
      "tree": "704f391120342d54f09e6aa49cce81e2b1b90cc8",
      "parents": [
        "9cf2c03053bbefde6178a4921151890f471e52d4"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 13:55:50 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] LenientAddressParser: a quoted pair in a domain literal is the escaped character alone\n\ncopyDomainLiteral copied a domain literal byte for byte up to the first\n\u0027]\u0027: the backslash of a quoted pair (obs-dtext, RFC 5322 section 4.4) was\nkept, and an escaped \"\\]\" closed the literal. So \"a@[1.2.3\\.4]\" gave the\ndomain \"[1.2.3\\.4]\" with the lenient parser, the default for From, To, Cc,\nReply-To and Sender, and \"[1.2.3.4]\" with the strict one, which reads a\nquoted pair as the character alone (RFC 5322 section 3.2.1).\n\nThe backslash of a quoted pair is now dropped and the escaped character\nkept, an escaped \u0027]\u0027 does not end the literal, and a line break after the\nbackslash is skipped, the quoted pair applying to the character after it,\nas in RawFieldParser.\n\nFixes #135\n"
    },
    {
      "commit": "9cf2c03053bbefde6178a4921151890f471e52d4",
      "tree": "5f6bee0ec4bffadd7df0344b805e37692b31aa87",
      "parents": [
        "d9feda8999be1343575cd5994067c7f2d0792f38"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 13 13:54:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 16 07:27:04 2026 +0100"
      },
      "message": "[FIX] ContentLocationFieldImpl: a form feed or vertical tab in the URL is kept\n\nThe strict parser removed the whitespace of the URL with the regex \"\\s\",\nwhich also matches form feed and vertical tab, while the lenient one\nremoves only SP, HTAB, CR and LF (CharsetUtil#isWhitespace). So\n\"Content-Location: a/\\fb\" gave \"a/b\" on the strict path (DefaultFieldParser,\nMaximalBodyDescriptor) and \"a/\\fb\" on the lenient DOM path.\n\nRFC 2017 drops linear whitespace, and FF/VT are control characters, not\nwhitespace: the strict parser now removes the same four characters as the\nlenient one.\n\nFixes #132\n"
    },
    {
      "commit": "d9feda8999be1343575cd5994067c7f2d0792f38",
      "tree": "4b9a0f730acd235f5ef7dc5394c99367f05afc69",
      "parents": [
        "3f642e4d4985be43c2e28d904978ce1436b9f687"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:15:10 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[PERF] BufferedLineReaderInputStream: release does not allocate a shift table\n\nSince 10ad0b96, release() replaced the shift table it hands back to the\nrecycler with a new int[256], so that a released stream never writes into\na table another stream now uses. That is 1 KB allocated for every stream\nreleased: 6 KB per parse of a multipart message, as the benchmarks showed.\n\nThe table is now dropped, and indexOf returns -1 on a released stream,\nwhose buffer is empty anyway, before the table is reached.\n"
    },
    {
      "commit": "3f642e4d4985be43c2e28d904978ce1436b9f687",
      "tree": "3e4c9d14747249cc283805c9f583f2424c397181",
      "parents": [
        "c984b2e1e3bc05874acf1ad8957b24edbecccbb0"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 12:16:56 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] RawFieldParser#skipComment: a backslash before a line break escapes what follows it\n\nIn a comment, a backslash escaped whatever byte came next, a bare CR or LF\nincluded. RawField#getBody deletes that line break (as MimeUtil#unfold),\nso the strict parsers read \"(\\\" CR \"(x)\" as the comment \"(\\(x)\", while\nskipComment let the backslash escape the CR, opened a nested comment at\nthe \"(\" and never closed it: \"; filename\u003da.exe\" was swallowed. With one\nbyte, any sender made a part\u0027s filename, parameters or Content-Language\ndiffer between the strict and the lenient parsers.\n\nAs copyQuotedContent already does, a line break after a backslash in a\ncomment is skipped and the quoted pair applies to the character after it.\n"
    },
    {
      "commit": "c984b2e1e3bc05874acf1ad8957b24edbecccbb0",
      "tree": "a52145cd9c7a7771dedfb66efc9c2924ceafe912",
      "parents": [
        "a026fedb17d069376e8c0deaa8b6f87c33797814"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 12:14:31 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] RawField#getBody: a fold right after the colon adds no leading space\n\ngetBody skipped one whitespace after the colon only when it was the byte\nright after it. For \"Subject:\" CRLF SP \"word\" that byte is the CR, so\nnothing was skipped, unfolding deleted the line break and kept the SP, and\nthe body read \" word\". MimeUtil#fold writes exactly that for a subject of\none word of 68 to 77 characters, which EncoderUtil does not encode: a\nsubject set by a JMAP user or a mailet came back with a leading space once\nthe message was written and parsed again, and so did the received fields\nof the mailers that fold that way.\n\nThe line break bytes right after the colon are now skipped before the one\nwhitespace, so \"Name:\" CRLF SP \"value\" reads as \"Name: value\" does.\n"
    },
    {
      "commit": "a026fedb17d069376e8c0deaa8b6f87c33797814",
      "tree": "586355f4d66f8ad0780361fd9647b01b0aec1e5d",
      "parents": [
        "462651a7d0fbb73dde9254e78acfa7cd9c0716d6"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 12:09:59 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] EncoderUtil: a parameter value holding \"\u003d?\" reads back unchanged\n\nencodeHeaderParameter wrote any value that is not a token as a quoted\nstring. The readers decode the encoded words of quoted parameter values,\nas many mailers write filename\u003d\"\u003d?utf-8?B?...?\u003d\": RawFieldParser decodes a\nquoted value starting with \"\u003d?\", and MimeParameterMapping a plain value\nholding one anywhere. So a filename or a parameter given to the builders\nas \"\u003d?utf-8?q?x?\u003d\" (a JMAP attachment name) was read back as \"x\" by every\nreader, from the built field and from the written message.\n\nWhen the field\u0027s parser decodes RFC 2231 (Content-Disposition), such a\nvalue is now written as an extended parameter, name*\u003dutf-8\u0027\u0027%3D%3F...,\nwhich is percent decoded and nothing more; in sections when it is too\nlong. Content-Type has no RFC 2231 reader, and its lenient reader only\ndecodes a value starting with \"\u003d?\": such a value is written as one B\nencoded word of the whole value. Values without \"\u003d?\" are written as\nbefore.\n"
    },
    {
      "commit": "462651a7d0fbb73dde9254e78acfa7cd9c0716d6",
      "tree": "13fc1f52d50e4f812b93027fdec6dcb9ba9ac028",
      "parents": [
        "5c10f60e63bc20c4e62c7d5b5d25b73ff384f24e"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 12:05:06 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] LenientAddressParser: a quoted local part is never decoded as an encoded word\n\nRawFieldParser#copyQuotedContent decodes a quoted string that starts with\n\"\u003d?\", for parameters such as filename\u003d\"\u003d?...?\u003d\", and LenientAddressParser\nread local parts through it: the angle-addr form called it directly, and\nthe local@domain form took the local part from parseValue, which decodes\ntoo. So \"\\\"\u003d?utf-8?q?x?\u003d\\\"@example.com\", which AddressFormatter writes for\na built Mailbox(\"\u003d?utf-8?q?x?\u003d\", ...), was read back as x@example.com.\nRFC 2047 section 5 forbids encoded words in an addr-spec and in a quoted\nstring.\n\nRawFieldParser gets copyQuotedContent(buf, cursor, dst, decodeEncodedWords),\nwhich local parts use with false, and a local part holding a quoted string\nis always read again as a local part. Display names and group names are\nstill decoded.\n"
    },
    {
      "commit": "5c10f60e63bc20c4e62c7d5b5d25b73ff384f24e",
      "tree": "aefcb67d8d842ba6f640faf44207ace98a8eabc2",
      "parents": [
        "6700011eec18b29aefd466980f4069894a65ab05"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 12:02:55 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] EncoderUtil#encodeAddressDomain: a domain literal cannot add recipients\n\nSince the fix of #77, a domain literal was written unchanged as long as it\nheld only dtext, which includes \",\" and \";\". LenientAddressParser stops its\nsearch for \"]\" at such a delimiter, so Mailbox(\"a\", \"[x,y]\") written as\n\"a@[x,y]\" was read back as two mailboxes, and \"team: a@[;];\" ended its\ngroup early. The domain comes from values given to the builders, or from\nan address DefaultAddressParser accepted.\n\nThe dtext characters a lenient reader takes for the end of an address\n(\",\", \";\", \"\u003c\", \"\u003e\", \"\\\"\", \"@\") are now dropped from a domain literal, as\nthey already are from any other malformed domain. \":\" (IPv6 literals) and\n\"(\" \")\", which the reader keeps inside a literal, are kept.\n"
    },
    {
      "commit": "6700011eec18b29aefd466980f4069894a65ab05",
      "tree": "b2e57983e90972ca91bd1332cd135b3e32968241",
      "parents": [
        "04ed8a6f7d9e5160bd266125b9c9a76aa3e5736f"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 11:59:18 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] DateTimeFieldLenientImpl: an out-of-range offset is not read as seconds\n\nThe obs-zone sections added by the fix of #24 are text fields of a lenient\nformatter, and a lenient text field that matches no name falls back to a\nnumber. So once appendOffset had rejected an offset such as \"+9900\" or\n\"+6357\", the obs-zone section read it as 9900 or 6357 seconds, and the\ndate was shifted by hours where the strict parser reads 99 hours: any\nsender could make the two parsers disagree on the sent date.\n\nThe obs-zone sections are now parsed strictly and match only a zone name;\nan offset nobody can read then gives no date. An offset without its sign\nright after the time (\"22:23:35.882 0000\", MIME4J-219), which only this\nfallback read, is now read explicitly as positive.\n"
    },
    {
      "commit": "04ed8a6f7d9e5160bd266125b9c9a76aa3e5736f",
      "tree": "200afc1e4d8123ad73dad19d2d008067f4757ac3",
      "parents": [
        "ccbf177e532edf722f133526b8f6b227b6d5aa18"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 11:55:07 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] DateTimeFieldLenientImpl: a zone after FWS or without space is honoured\n\nEach zone section of RFC_5322 began with a single space literal, and the\nparse ignored where it stopped. A zone after two spaces or a tab (valid\nRFC 5322 FWS: \"09:55:06  -0600\"), an obs-zone with no space before it\n(\"1:2EDT\", allowed by RFC 5322 4.3), or an offset read only in part (the\n\"+11\" of \"+1178\") was dropped with the rest of the value, and the time was\nread as GMT or with a partial offset: any sender could have the stored\nsent date shifted by hours. This is what the fix of #24 left open.\n\nRuns of white space now read as one space before parsing, the space\nbefore an obs-zone is optional, and a value whose numeric zone is left\nunread, in whole or in part, gives no date rather than a shifted one.\nOther trailing text, such as a comment, is still tolerated.\n"
    },
    {
      "commit": "ccbf177e532edf722f133526b8f6b227b6d5aa18",
      "tree": "be8f317c3efe996b0e3d6ada68450885e75632f9",
      "parents": [
        "527a881ec43d2ae5e445a5828120c565c2890386"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 11:50:58 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] QuotedPrintableOutputStream: a dash never starts an output line\n\nThe encoder escaped only \"\u003d\" and \".\" among the printable characters, and\nwrote the byte that follows its own soft break as is. So a \"-\" could start\nan output line, after a line break of the content or after a soft break,\nand the encoded body could hold \"--\u003cboundary\u003e\". DefaultMessageWriter\nwrites the parts of a multipart between delimiters of the original\nboundary: a quoted-printable part whose text had \"--b\" at column 76, or\nwhose decoded text held a delimiter line (\"\u003d2D-b\", or a soft break inside\n\"--b\u003d\" CRLF \" inline\"), was written with that delimiter, and the\nre-parsed message had parts the first parse never saw, with headers the\nsender chose. Any sender could trigger it, and James re-serialises the DOM\nin many places.\n\nA \"-\" that would be the first byte of an output line, including the line\nopened by a soft break, is now written as \"\u003d2D\" (RFC 2045 6.7 rule 1), so\nthe quoted-printable output never holds a delimiter, whatever the\nboundary (RFC 2046 5.1.1). The expected outputs of six example messages\nchange accordingly: they decode to the same content.\n"
    },
    {
      "commit": "527a881ec43d2ae5e445a5828120c565c2890386",
      "tree": "5f76753348dbb1625ee0ca1b7bb35a6cdd246155",
      "parents": [
        "46b4d91a3ec6fdd0ab038cd5bead848ab5ec18bc"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 10:58:30 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] CipherStorageProvider: encrypt with segmented AES-256/GCM\n\nThe provider defaulted to \"Blowfish\" and built its ciphers from the bare\nalgorithm name, which SunJCE reads as \u003calg\u003e/ECB/PKCS5Padding: identical\nplaintext blocks gave identical ciphertext blocks, no IV was used, and\nnothing authenticated the ciphertext. Whoever could read the spool learnt\nthe structure of each attachment, which the provider exists to hide, and\nwhoever could write it changed the bytes the parser read back undetected.\n\nEach storage now gets a random AES-256 key and a random nonce prefix, kept\nin memory only. The data is cut into 64 KB segments, each encrypted and\nauthenticated with AES/GCM under a nonce made of that prefix, the segment\nnumber and a last-segment flag: a modified, reordered or truncated content\nfails the read with an IOException before any altered byte is returned.\nSegments keep the memory of a read bounded, which a single GCM operation\nthrough CipherInputStream would not, as it buffers the whole part to check\nits tag. The two-argument constructor accepts only \"AES\" and\n\"AES/GCM/NoPadding\" and refuses other algorithms with an\nIllegalArgumentException, ECB and bare names such as \"Blowfish\" included.\n"
    },
    {
      "commit": "46b4d91a3ec6fdd0ab038cd5bead848ab5ec18bc",
      "tree": "964b5dc6417c6a23c90006589b3fcc3cc79d011e",
      "parents": [
        "ffa0dbd2f52695222b0bd53e4d5f316e0a42ef6e"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 10:53:19 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] TempFileStorageProvider: a deleted temporary file leaves nothing registered\n\nEach spooled part called File.deleteOnExit(), which adds its path to the\nstatic set of java.io.DeleteOnExitHook. That set is read only at shutdown\nand never shrinks, even once TempFileStorage.delete() removed the file. As\nthe default provider spools every part of 1 KB or more, a long-running\nservice parsing mail through StorageBodyFactory kept one path per part in\nthe heap until the JVM stopped: a sender sending many messages of small\nparts, or ordinary traffic over time, made the heap grow without bound.\nFiles that could not be deleted also stayed in an unbounded static set,\nretried by every later delete() under one global lock.\n\nThe provider now keeps the files not deleted yet in a concurrent set that\ndelete() removes them from, and one shutdown hook deletes whatever is left.\nFailed deletions are retried by later deletions without a global lock, up\nto MAX_PENDING_DELETIONS files; the others are left to the shutdown hook.\n"
    },
    {
      "commit": "ffa0dbd2f52695222b0bd53e4d5f316e0a42ef6e",
      "tree": "4a7a787da1c9de32c14bbf033de454f114f83e55",
      "parents": [
        "2242af35660d3a593053735f949e8d25ed51ddec"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 10:52:26 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] BasicBodyFactory: a disposed body no longer reaches its recycled buffer\n\nStringBody3 and BinaryBody3 hand their buffer back to the per-thread\nByteArrayOutputStreamRecycler on dispose() but kept using it: the next\nbody buffered on that thread, often the next message and another user\u0027s\nmail, took the same UnsynchronizedByteArrayOutputStream, and from then on\ngetInputStream, getReader, writeTo and size of the disposed body returned\nthat other content without error. copy() after dispose() brought the\nreleased buffer back to a count of 1, so the copy and the next owner\nshared one buffer.\n\nA disposed body now drops its buffer and throws IllegalStateException\n(\"body has been disposed\") from these methods and from copy(), as\nTempFileStorage already does. ByteArrayOutputStreamRecycler.Wrapper refuses\nretain() and getValue() once released, and a second release no longer\ncounts below 0. The other recycled holders drop their arrays on release\ntoo: RecycledByteArrayBuffer falls back to plain arrays if it is still\nappended to (a field read after stop() wrote into the recycled array), and\nBufferedLineReaderInputStream reads nothing more. Streams opened before\ndispose() keep their content.\n"
    },
    {
      "commit": "2242af35660d3a593053735f949e8d25ed51ddec",
      "tree": "adb72188046c683b857034dab793474cbd562a30",
      "parents": [
        "b256a924f9e824b5ef9ccb9dd04f90e2c4568bdf"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 10:49:31 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] MimeEntity: a header line is not buffered beyond maxHeaderLen\n\nreadRawField read each header line whole, then appended it to the field\nbuilder, which only then checked maxHeaderLen. Under MimeConfig.PERMISSIVE,\nmaxLineLen is -1, so nothing bounded that read: a header made of \"X-A: \"\nfollowed by tens of MB without a line feed was buffered entirely, in a\nByteArrayBuffer that grows by doubling (2 to 3 times its size at peak),\nbefore MaxHeaderLengthLimitException was thrown. The headers of body parts\nwere read the same way. James parses every stored message with PERMISSIVE,\noften in parallel, so a burst of such messages could exhaust the heap,\ndespite the javadoc promising the profile stays bounded.\n\nLineReaderInputStream gets readLine(dst, maxLen), which stops once dst\nholds more than maxLen bytes and no line feed; BufferedLineReaderInputStream,\nMimeBoundaryInputStream and LineReaderInputStreamAdaptor implement it, and\nthe default implementation keeps the old behaviour for other subclasses.\nreadRawField reads with maxHeaderLen + 1 as the bound, so it only cuts\nshort a line no field of maxHeaderLen bytes can hold, and then throws\nMaxHeaderLengthLimitException. Fields of exactly maxHeaderLen bytes are\nstill accepted, and body lines stay unbounded under PERMISSIVE. The javadoc\nof PERMISSIVE now says what is bounded.\n"
    },
    {
      "commit": "b256a924f9e824b5ef9ccb9dd04f90e2c4568bdf",
      "tree": "20d0020fa1db4d7d554eab7a5b9dba2db3db06d6",
      "parents": [
        "a3a9ec2e578e5fd05bf4bbb58e904237f09e9cb5"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 10:46:39 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 12 16:53:46 2026 +0100"
      },
      "message": "[FIX] JavaCC token managers: the comment-nesting counter is per parser\n\nAddressListParser, ContentTypeParser, ContentDispositionParser and\nDateTimeParser set STATIC\u003dfalse, and callers build a new parser for each\nparse, but their TOKEN_MGR_DECLS declared \"static int commentNest\". So\nthe JVM had one counter per grammar, shared without synchronisation by\nevery thread. A nested comment such as \"(x (y) z)\" leaves NESTED_COMMENT\nonly when that counter comes back to exactly 0: a parse running on\nanother thread at the same time made it skip 0, and the lexer swallowed\nthe rest of the field and failed at EOF, or reach 0 too early, and the\nlexer read comment text as tokens. Any sender could thus make the strict\nparse of other messages, often other users\u0027 mail, fail or return wrong\nmailboxes, MIME type, parameters or date, and the same happened under\nordinary concurrent load.\n\nThe counter is now an instance field of the token manager, as\nStructuredFieldParser, ContentLanguageParser and MimeVersionParser\nalready declared it. A test parses nested comments with the four parsers\non concurrent threads, and another fails if a grammar with STATIC\u003dfalse\ndeclares a static field in TOKEN_MGR_DECLS.\n"
    },
    {
      "commit": "a3a9ec2e578e5fd05bf4bbb58e904237f09e9cb5",
      "tree": "ca9657b70dd37b51526f19fa2adea9a16cfce83c",
      "parents": [
        "2dee0e7a2ef4df0f246cdacbd727e359ac61ba89"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 10 21:52:34 2026 +0100"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Tue Nov 10 21:52:34 2026 +0100"
      },
      "message": "[UPGRADE] guava 33.7.1-jre -\u003e 33.7.2-jre (#140)\n\nguava 33.7.1 is affected by CVE-2026-102554 (unthrottled allocation when\ndeserializing CompactHashMap/CompactHashSet/MapMakerInternalMap)."
    },
    {
      "commit": "2dee0e7a2ef4df0f246cdacbd727e359ac61ba89",
      "tree": "0cf424ed64598ef57157f71797016914ecc1abcd",
      "parents": [
        "b0c40752f2fe61515983cfce705476686bca0324"
      ],
      "author": {
        "name": "OldTruckDriver",
        "email": "stayintouchwithdrq@gmail.com",
        "time": "Sun Nov 08 02:19:30 2026 +1100"
      },
      "committer": {
        "name": "Rene Cordier",
        "email": "rene.cordier@gmail.com",
        "time": "Mon Nov 09 09:17:28 2026 +0700"
      },
      "message": "MIME4J-335 DecoderUtil: treat a null DecodeMonitor as SILENT\n\ndecodeEncodedWords(String, Charset) passed null as the DecodeMonitor to\nthe four-argument implementation. The monitor is dereferenced by\nDecoderUtil.monitor(...) and by the decoding streams, so any encoded\nword that could not be decoded, such as a base64 word without padding,\nan unknown encoding or malformed Q text, made this overload throw a\nNullPointerException.\n\nA null monitor is now replaced by DecodeMonitor.SILENT at the entry of\nthe four-argument implementation. This also covers callers that pass\nnull explicitly to the three- and four-argument overloads, and makes the\nCharset overload return what the DecodeMonitor.SILENT overload returns.\n"
    },
    {
      "commit": "b0c40752f2fe61515983cfce705476686bca0324",
      "tree": "74f4ce1446a37ab777e84a628506d9520b4bc65e",
      "parents": [
        "994242bd2a27c6ea8c64a93407b47c23ba1ea3ff"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 12:30:32 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] CharsetUtil#forEncoding: text its charset cannot encode takes UTF-8\n\nSince 5a754cfb the builders rejected a text its charset cannot encode. But\nthe charset and the text can both come from a received part: decoding\nreplaces an undefined byte with U+FFFD, which US-ASCII, ISO-8859-x or\nwindows-125x cannot encode. So any sender could make code that rebuilds a\npart in its own charset (footer, quote, forward) fail with an\nIllegalArgumentException, with one 8-bit byte in a part labelled US-ASCII.\n\nSuch a text now takes UTF-8 and the part is labelled with it, as already\ndone for a decode-only charset or when no charset is given: the text is\nneither altered nor rejected. Only a lone surrogate, which no charset\nencodes, is still rejected. StorageBodyFactory, which still wrote \"?\", now\ndoes the same.\n"
    },
    {
      "commit": "994242bd2a27c6ea8c64a93407b47c23ba1ea3ff",
      "tree": "ab597d0e530f06d7cab97a9c2729522aa3ed19fe",
      "parents": [
        "e9f833f1346b1727f4089893538463f4ae42d75b"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 12:28:44 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] MimeEntity: closing a body of an encapsulated message does not close its parent\n\nMimeStreamParser closes each body stream, and that close went down to what\nthe encapsulated entity reads: the decoder of the message/rfc822 part, or\nthe caller\u0027s stream. Since dab2f50f, leaving T_BODY reads what is left of\nthe content when maxContentLen is set, so the read hit the closed stream.\nUnder PERMISSIVE, parse() failed with an IOException on a base64 or\nquoted-printable message/rfc822 whose body the handler left unread, or\nholding another message/rfc822, and read the caller\u0027s stream after closing\nit without an encoding.\n\nThe encapsulated entity now reads through a stream that ignores close(),\nas MimeBoundaryInputStream#close already does inside a multipart. Its parent\ncloses the decoder once it has read its content. A body of a message/rfc822\ntherefore no longer closes the caller\u0027s stream.\n"
    },
    {
      "commit": "e9f833f1346b1727f4089893538463f4ae42d75b",
      "tree": "4452fec48581f0938780fdbd7a57f3e7897b0ccc",
      "parents": [
        "040d3a1595a8631b50cbc6d16c7831cd16ce0bad"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 05:34:39 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] DefaultMessageWriter: a nested multipart does not close an enclosing one\n\nA nested multipart that declares the boundary of its parent can only hold\na preamble: the parser ends it at the first delimiter of the parent. The\nwriter still gave it a closing delimiter, which the next parse read as the\nend of the parent: the following body parts became epilogue, and each\nfurther write added one more delimiter.\n\nSuch a multipart is now written as its preamble only, what the first parse\nread. A boundary declared by any enclosing multipart counts, through a\nmessage/rfc822 part too, unless base64 hides the delimiter.\n"
    },
    {
      "commit": "040d3a1595a8631b50cbc6d16c7831cd16ce0bad",
      "tree": "aa0b47e01ae81fb4c73cc435caceac1c6ce12746",
      "parents": [
        "dbc108d8a34e3ff1f0e76520df34c67a60dac6e1"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 05:32:38 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] maxContentLen: a body part does not depend on how its source splits reads\n\nA body part left the rest of its content to its parent, which drained it\nthrough a new limited stream counting from 0. So only what a decoder had\nnot read ahead was counted: with data after a base64 pad, a part over the\nlimit was accepted or rejected depending on the size of the reads, and a\nmessage/rfc822 part on its position in the parser\u0027s buffer.\n\nA body part now drains its content itself, through the limited stream that\nalready counted those bytes. An encapsulated message limits its entities\none by one, so its own bodies are drained the same way, and what is left\nof it (data after the pad, an epilogue) is read and discarded without a\nlimit: the encoded form of a message is not bounded as a whole.\n"
    },
    {
      "commit": "dbc108d8a34e3ff1f0e76520df34c67a60dac6e1",
      "tree": "2a9a38dfab17109be9767420abe1d1651d105220",
      "parents": [
        "bf7a8fb48f23542548719e324d5cd9a912d34bc3"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 05:30:15 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] DOM builders: do not silently replace the text a charset cannot encode\n\nThe text of a TextBody is encoded with CodingErrorAction.REPLACE, so a\ntext its charset cannot encode was written with \"?\": the sent body\ndiffered from the built one, with no error and no log.\n\nThe builders and BasicBodyFactory now check the text:\n- a charset chosen by the caller has to encode it, otherwise\n  IllegalArgumentException. So does a lone surrogate, in any charset;\n- when the charset is only a default, a text that the default cannot\n  encode takes UTF-8, and the part is labelled with it.\n"
    },
    {
      "commit": "bf7a8fb48f23542548719e324d5cd9a912d34bc3",
      "tree": "2b8ae2d69cb15ed9d19b18f7ee8cf370d20732ad",
      "parents": [
        "d8aa672222b1752e6fd1ad9485eccc5b09e329a8"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 05:27:40 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] RawFieldParser: a quoted pair in a quoted charset stands for the escaped character\n\nThe backslash was dropped for the boundary only, so charset\u003d\"iso-8859\\-1\"\ngave an invalid charset name to the lenient parsers: the default DOM fell\nback on the JVM charset and MimeTokenStream#getReader threw, whereas the\nstrict parser honoured iso-8859-1. A charset name cannot hold a backslash\n(RFC 2978), so its quoted pairs are now unescaped too. Encoded-word\ndecoding and the other parameters do not change.\n"
    },
    {
      "commit": "d8aa672222b1752e6fd1ad9485eccc5b09e329a8",
      "tree": "8673dac9ef60bfe0e9734c2d49818914f17d47fc",
      "parents": [
        "cea1f802d0d3d212fb0a528655e3b30db7291970"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 05:26:45 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] RawField#getBody: delete CR and LF before decoding, as RawFieldParser\n\ngetBody decoded the bytes as UTF-8 and then unfolded the text, whereas\nRawFieldParser leaves line breaks out of the bytes it decodes. A bare CR\ninside a multi-byte sequence therefore split it for the strict parsers\nonly: filename\u003d\"caf\u003cC3\u003e\u003cCR\u003e\u003cA9\u003e.txt\" read caf\u003cU+FFFD\u003e\u003cU+FFFD\u003e.txt strict\nand café.txt lenient. getBody now deletes the line breaks from the bytes\nfirst, so that both paths agree.\n"
    },
    {
      "commit": "cea1f802d0d3d212fb0a528655e3b30db7291970",
      "tree": "865b6fb0530ab8b83aee346a242473d24b8a9330",
      "parents": [
        "5c664e85939767c586b3eaf9c532f24adf532f36"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 05:25:51 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] LineReaderInputStreamAdaptor: skip(n) never skips more than n\n\nskip() asked for the whole buffer on every read. After a short read, for\nexample at the edge of the parser\u0027s buffer or at the limit of a boundary\nstream, what was left to skip was smaller than the buffer, so skip(n)\nconsumed and returned more than n: up to 8191 extra bytes were dropped.\nEach read now asks for what is left to skip.\n"
    },
    {
      "commit": "5c664e85939767c586b3eaf9c532f24adf532f36",
      "tree": "976b51c8f0daf3788278fd47603db22ea03e770e",
      "parents": [
        "319fe5425e3d0494bdb7733eedc53b41c8589f0f"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Thu Nov 05 05:25:13 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] Strict parsing: stop() at the start of an entity is not an invalid header\n\nSince 23c513fe a header cut short by stop() no longer reports\nHEADERS_PREMATURE_END, but nextField went on building a field from what\nit had. With no field read yet the builder is empty, and the failure was\nreported as INVALID_HEADER, which strict parsing turns into a\nMimeParseEventException: stop() at T_START_MESSAGE, T_START_HEADER or\nT_START_BODYPART failed on valid mail. So did a stop() on a field whose\nread-ahead line is malformed, a line the caller had stopped before.\n\nA field that fails after stop() is not reported: the header ends there.\n"
    },
    {
      "commit": "319fe5425e3d0494bdb7733eedc53b41c8589f0f",
      "tree": "b775187b5da9c351fb52f678ea0c72bad45f56fc",
      "parents": [
        "eacd72f1cbafa8b8c0529cfeb5f8307d43091cad"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 19:41:37 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] RawFieldParser: a bare CR or LF is deleted, as MimeUtil#unfold does\n\nThe strict parsers read the field body through MimeUtil#unfold, which\ndeletes every CR and LF. RawFieldParser instead took a CR or LF as a word\nseparator, and kept a quoted pair \"\\\" + bare CR as an escaped CR. So\n\"multi\u003cCR\u003epart/mixed\" was multipart for the strict parser only, and\nContent-Disposition, Content-Language, MIME-Version and Content-Location\nvalues differed between the strict and lenient parsers in the same way.\n\nRawFieldParser now deletes a CR or LF wherever it is: only the WSP of a\nfold separates words, and in a quoted pair the backslash applies to the\ncharacter after the line break. This supersedes a8ed914c, which kept the\nCR after a backslash: the strict parsers never did.\n"
    },
    {
      "commit": "eacd72f1cbafa8b8c0529cfeb5f8307d43091cad",
      "tree": "d5ed30b022af6512ceb55966cd14391ec3968ddd",
      "parents": [
        "5080ca7c1f35eb3cb8711e44cd5a5a9e35b9670c"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 19:36:16 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] Strict parsing: stop() in a header or before a message/rfc822 body is not an error\n\nAfter stop(), the truncated buffer ends the header, and readRawField\nreported it as HEADERS_PREMATURE_END, which strict parsing turns into a\nMimeParseEventException. And at the end of the header of a message/rfc822\nentity, the parser still created the encapsulated entity: created after\nstop(), it was not stopped and failed on its empty header. So a strict\nDefaultMessageBuilder#parseHeader failed on any message/rfc822 mail.\n\nA stopped entity no longer reports the end of its header as premature,\nand no longer recurses into an encapsulated message: its body is empty.\n"
    },
    {
      "commit": "5080ca7c1f35eb3cb8711e44cd5a5a9e35b9670c",
      "tree": "3ce52acab1b220028b0acecc6ade8e1cdd073e5e",
      "parents": [
        "a18cb891ef57e587a378cd70fdbf924bbda85384"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 19:34:39 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] stop() in a multipart: never start a new body part\n\nAfter stop(), MimeEntity still opened the next body part whenever the\npart stream had not reached the last delimiter. Two paths reached it:\n\n- with malformedHeaderStartsBody, a field cut short by stop() unread the\n  line read ahead into the truncated buffer. Reads were refused, so the\n  part stream never reached eof() and an empty body part was started\n  again and again, forever.\n- after a preamble or body part was read to its end, the delimiter had\n  already been skipped, so one empty body part followed stop(), and\n  strict parsing failed on its empty header.\n\nOnce stopped, the multipart now ends right away (T_END_MULTIPART), and\nan unread into a truncated buffer drops the bytes like the rest of it.\n"
    },
    {
      "commit": "a18cb891ef57e587a378cd70fdbf924bbda85384",
      "tree": "5d47e39fb7e2b0cd80a42cddc558e149ad6921d4",
      "parents": [
        "72e2ddd7643832fff3e1e136e8fa5f0c9af9a65a"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:14:06 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] Fields: keep a long Content-Disposition parameter within the line limit\n\nEncoderUtil#encodeHeaderParameter wrote a parameter as one token or\nquoted string whatever its length, and folding never breaks a run\nwithout whitespace: a long filename or Content-Type parameter gave a\nheader line over 998 octets that DefaultMessageBuilder rejects. A new\nencodeHeaderParameter(name, value, rfc2231Sections) splits such a\nparameter into RFC 2231 sections (percent encoded in UTF-8 when not\nASCII), which the Content-Disposition parser decodes, and rejects it\nwith an IllegalArgumentException otherwise: no Content-Type parser\ndecodes RFC 2231 sections, so Fields.contentType rejects such a value.\n\nFixes #105\n"
    },
    {
      "commit": "72e2ddd7643832fff3e1e136e8fa5f0c9af9a65a",
      "tree": "39a6b1e8e6805ee3928f64d271bb3259515a31c4",
      "parents": [
        "b90149fb8520ac6d7839a45625d7c769c4266a0c"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:04:24 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] DefaultMessageWriter: delimit a multipart with the Content-Type that describes the body\n\nSince c9847207 the parser and AbstractEntity skip a leading Content-Type\nthat gives no mime type or is multipart without boundary, and use the\nnext usable one. DefaultMessageWriter still read the first Content-Type\nto find the boundary, so writing such a parsed Multipart threw\n\"Multipart boundary not specified\". The writer now picks the field with\nthe same rule, through a helper shared with AbstractEntity.\n\nFixes #104\n"
    },
    {
      "commit": "b90149fb8520ac6d7839a45625d7c769c4266a0c",
      "tree": "2363ae0365ed92d2aa25cc2ddc4baadfc86adbd6",
      "parents": [
        "fbfbf0e2d6d41cf32203a72dc8fd76ebcf46a944"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:12:49 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] RawFieldParser: a quoted pair in a quoted boundary stands for the escaped character\n\ncopyQuotedContent keeps the backslash of a quoted pair before any\ncharacter other than a quote or a backslash, so boundary\u003d\"a\\bc\" was read\nas a\\bc by the lenient parsers and MimeTokenStream, while the strict DOM\nparser reads abc: the two split the multipart on different delimiters.\nThe boundary already has its own path (no encoded-word decoding); it now\ndrops the backslash of every quoted pair too, as RFC 2046 bchars hold no\nbackslash. Other parameters are unchanged.\n\nFixes #108\n"
    },
    {
      "commit": "fbfbf0e2d6d41cf32203a72dc8fd76ebcf46a944",
      "tree": "161d4a56ab2dc8ae6d6e20bc480e6c4c222db027",
      "parents": [
        "005f0fe79767da149b346ee5784b7e8b107a2d96"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:10:07 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] FallbackBodyDescriptorBuilder: take the charset of the first Content-Type when none is usable\n\nFallbackBodyDescriptorBuilder parses every Content-Type field until one\ngives a usable type, and each one replaced the charset, so with no usable\nfield the charset came from the last one. DefaultBodyDescriptorBuilder\nkeeps the first field in that case, and core and DOM decoded the body with\ndifferent charsets. A later field now replaces the charset only when its\ntype is accepted, as DefaultBodyDescriptorBuilder replaces the field.\n\nFixes #107\n"
    },
    {
      "commit": "005f0fe79767da149b346ee5784b7e8b107a2d96",
      "tree": "adfc0a3cd58fa57d3cb06f97ca7cb0502d5f45f9",
      "parents": [
        "e377ed69f6754ebb4ed298e43486736047e40de1"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:06:32 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] StructuredFieldParser: drop the backslash of a quoted pair followed by text in a quoted string\n\nThe quoted pair is a MORE rule whose action removes the backslash from\nthe token manager\u0027s image. STRING_CONTENT and FOLD had no action, so\nJavaCC built their image from the raw input, MORE prefix included: the\nbackslash came back (\"a\\\"b\" read as a\\\"b, the lenient parser gives a\"b),\nand a quoted pair just before a fold was dropped. Both tokens now take\ntheir image from the edited buffer, the fold keeping only the text before\nthe line break.\n\nFixes #106\n"
    },
    {
      "commit": "e377ed69f6754ebb4ed298e43486736047e40de1",
      "tree": "5bca91361962180a7422819f02e65e1abef308ca",
      "parents": [
        "4a0f8218c61fdde09316a4c9f97c10630fd7b50e"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:10:48 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] QuotedPrintableOutputStream: encode whitespace that ends the content\n\ncompleteEncoding() wrote a pending space or TAB unencoded, but the end of\nthe content ends an encoded line, so the decoder drops that trailing\nwhitespace (RFC 2045 6.7 rule (3)) and a body ending with a blank lost it\non the next parse. The last space or TAB is now written as \u003d20 / \u003d09.\n\nFixes #110\n"
    },
    {
      "commit": "4a0f8218c61fdde09316a4c9f97c10630fd7b50e",
      "tree": "a31b9c22090620365cca5e46358104e820f7eabe",
      "parents": [
        "726042ff212a9056ce41d620367314976f46398d"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:08:55 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] QuotedPrintable: keep lone CR and LF through a decode / encode round trip\n\nOn a malformed escape such as \"\u003d2\" CRLF, QuotedPrintableInputStream wrote\nthe byte after the escape as data even when it was a CR, so the CRLF became\n\"\\r\\r\\n\". The text-mode QuotedPrintableOutputStream then lost a CR followed\nby another CR, a pending CR after a pending blank, and wrote any other lone\nCR or LF unencoded, which the decoder turns into CRLF (or rejects in strict\nmode). DefaultMessageWriter therefore changed such bodies. The decoder now\nleaves only \"\u003d\" and the first byte as is and handles the third one as usual;\nthe encoder keeps only CRLF as a line break and escapes any other CR or LF\n(\u003d0D, \u003d0A), as RFC 2045 6.7 rule (4) requires.\n\nFixes #109\n"
    },
    {
      "commit": "726042ff212a9056ce41d620367314976f46398d",
      "tree": "4b0b453d4ca82cfafd7d87133ea16a97c86aab2a",
      "parents": [
        "1c7690c54ecce3f3da27160e71a63495060c5539"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:10:13 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] MimeBoundaryInputStream: scan the buffered bytes before reading more\n\nThe constructor always read the underlying stream first, before looking\nfor the delimiter in the bytes already buffered. For a nested multipart\nread in one go, the enclosing part was then already exhausted: in strict\nmode, the extra read of an enclosing part without a closing delimiter\nthrew \"Unexpected end of stream\" before the nested preamble, while small\nreads reported the preamble and the next part first. The constructor now\nscans the buffered bytes first. That read also told, as a side effect,\nwhether the source ends right after a delimiter line: skipBoundary now\nreads ahead when nothing is buffered after the delimiter, so that eof()\nstays the same.\n\nFixes #111\n"
    },
    {
      "commit": "1c7690c54ecce3f3da27160e71a63495060c5539",
      "tree": "371198c289cfa7bef9611e02d4cd1a6ccf47fd87",
      "parents": [
        "76e0063c3fbf3db34252c3fd38d19c2b87a7c3d1"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:04:22 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] MimeTokenStream: stop() also stops the nested entities\n\nstop() only truncated the root entity. A nested multipart or message\nreads its header and body from its own buffer, refilled from its parent:\nafter stop() it went on reporting fields, parts and bodies as if nothing\nhappened. stop() now stops every open entity, so only the end events of\nthe open entities follow.\n\nFixes #102\n"
    },
    {
      "commit": "76e0063c3fbf3db34252c3fd38d19c2b87a7c3d1",
      "tree": "e99baa3cdf4613076a49eee40d341d3de295c041",
      "parents": [
        "4383a37b469fd2861c3b3143d10ae628f103ae7b"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 15:02:32 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[TEST] Strict parseHeader accepts a valid multipart (#103)\n\nstop() before the body of a multipart made strict parsing throw\n\"Unexpected end of stream\". Fixed by the previous commit, which treats\nthe truncation stop() asks for as the end of the multipart rather than a\npremature end. This adds the regression test of the issue: both inputs\nof the ticket, DEFAULT and PERMISSIVE, read whole and in chunks.\n\nFixes #103\n"
    },
    {
      "commit": "4383a37b469fd2861c3b3143d10ae628f103ae7b",
      "tree": "15048c8a4af4fe8318b1eb2c90cb6a7fdf33f4f2",
      "parents": [
        "c5a88f2c1b7b22e406fc3fbaee9701dbec615e3a"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 14:56:55 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[FIX] stop() inside a multipart: read nothing more and end the multipart\n\nstop() truncates the root buffer, but only its read() and readLine()\nchecked the truncation. BufferedLineReaderInputStream#fillBuffer, used\ndirectly by MimeBoundaryInputStream, kept reading from the caller\u0027s\nstream. The boundary stream then found the next delimiter in the refill,\nits part ended without eof nor last part, and MimeEntity opened the next\npart at the same position: new body parts after stop(), and with some\nreads an endless loop (DefaultMessageBuilder#parseHeader never returned).\n\n- BufferedLineReaderInputStream#fillBuffer returns -1 once truncated.\n- MimeBoundaryInputStream treats a truncated buffer as the end of the\n  stream: eof is set, nothing is skipped, so the multipart ends with\n  T_END_MULTIPART.\n- MimeEntity does not report a part cut by stop() as a premature end,\n  which strict parsing turned into an error.\n\nFound by twake-fuzz, scenarios 30-stop-contract and\n31-parse-header-matches-message.\n"
    },
    {
      "commit": "c5a88f2c1b7b22e406fc3fbaee9701dbec615e3a",
      "tree": "dff70ee2dc5040b940d5cf1f6b9617b4127244aa",
      "parents": [
        "d4d13f467ffacc323c1c8af9e7b9fc92c2a71055"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 14:54:37 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[PERF] AddressListParser: disable JavaCC error reporting, quadratic on long lookaheads\n\naddress() starts with an unbounded syntactic lookahead for addr_spec. On a\nvalue such as a\"b\".a\"b\"... it scans the whole value, fails, falls back to\nphrase() and throws early. generateParseException() then ran\njj_rescan_token(), which walks the token list from the current token for\nevery scanned token: O(n^2), 2-3 s for a 32 KB To header under the fuzzer.\n\nWith ERROR_REPORTING\u003dfalse JavaCC generates neither the rescan nor the\nexpected-token bookkeeping. The ParseException message still gives the\nposition and the offending token, but no longer lists the expected tokens.\n\nFound by twake-fuzz, scenario 33-header-value-cost-scaling.\n"
    },
    {
      "commit": "d4d13f467ffacc323c1c8af9e7b9fc92c2a71055",
      "tree": "3368873b183593a171d901e65153963dc52fac02",
      "parents": [
        "23a562ac629abf7527034e8e9e051675873e5baf"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 14:53:28 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[PERF] DecoderUtil: stop at the first \"\u003d?\" that cannot complete an encoded word\n\nWhen the search for the charset \u0027?\u0027, the encoding \u0027?\u0027 or the closing \"?\u003d\"\nfailed, decodeEncodedWords advanced by two characters to the next \"\u003d?\"\nand searched to the end of the value again: O(k*n) for k \"\u003d?\" with no\n\"?\u003d\", in Subject, display names and parameter values.\n\nThe three searches only move forward with the start of the pattern, so\nonce one fails no later \"\u003d?\" can complete an encoded word. Copy the rest\nof the value and stop. The output is unchanged.\n\nFound by twake-fuzz, scenario 33-header-value-cost-scaling.\n"
    },
    {
      "commit": "23a562ac629abf7527034e8e9e051675873e5baf",
      "tree": "806e3c370ea3ff7d00de1820364277fb415442bb",
      "parents": [
        "c9847207a5aa3ce2526bfee78edf3dbae9c22e35"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Wed Nov 04 14:52:39 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Nov 06 07:44:45 2026 +0100"
      },
      "message": "[PERF] LenientAddressParser: do not rescan for \u0027]\u0027 after each unclosed \u0027[\u0027\n\nparseDomain called copyDomainLiteral at every \u0027[\u0027. When no \u0027]\u0027 follows,\nthe scan ran to the next delimiter or the end of the value and was\nrepeated for the next \u0027[\u0027: O(k*n) for k unclosed brackets, over 16 s for\na 32 KB To header under the fuzzer.\n\ncopyDomainLiteral now returns where a failed scan stopped. Every \u0027[\u0027\nbefore that index is unclosed as well, so parseDomain copies it as\ncontent without scanning again.\n\nFound by twake-fuzz, scenario 33-header-value-cost-scaling.\n"
    },
    {
      "commit": "c9847207a5aa3ce2526bfee78edf3dbae9c22e35",
      "tree": "bec7893a0405c36b7ecf5aeba962dabc4a1d143f",
      "parents": [
        "9b76561f927920e34da53e4cb97af1cab69b1506"
      ],
      "author": {
        "name": "Rene Cordier",
        "email": "rcordier@linagora.com",
        "time": "Tue Nov 03 14:31:45 2026 +0700"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 03 16:19:51 2026 +0300"
      },
      "message": "fix: ignore a malformed leading duplicate Content-Type / Content-Transfer-Encoding header so core and DOM agree (PR #137)\n"
    },
    {
      "commit": "9b76561f927920e34da53e4cb97af1cab69b1506",
      "tree": "bd1b67b4e110077b75ef19fb17b967c88fcc0e4e",
      "parents": [
        "a6d5dce29a7eb21ab0ba84c958fda5e59cf518fa"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 15:26:13 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 03 16:19:51 2026 +0300"
      },
      "message": "[FIX] Strict Content-Language parser: a subtag needs its hyphen, without whitespace\n\nThe alternative of the subtag loop held a bare ALPHANUM, the \"-\" being\nin the first alternative only, and its action still added a \"-\": \"en\n4t\" read as en-4t. Whitespace is a special token, so \"zh-Hant- in\" read\nas zh-Hant-in. A subtag is now \"-\" then ALPHA or ALPHANUM, and\nwhitespace around that \"-\" is a parse error.\n"
    },
    {
      "commit": "a6d5dce29a7eb21ab0ba84c958fda5e59cf518fa",
      "tree": "587f5a4b044f1b130e5c3baf7db467ca3922720a",
      "parents": [
        "c8ff400ceb7d28ee2cedb7c87fc190ff400bb9b3"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 15:25:27 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 03 16:19:51 2026 +0300"
      },
      "message": "[FIX] Strict Content-Language and MIME-Version parsers: parse the whole value\n\nBoth stopped at the first token that did not continue the grammar and\nreported the valid prefix without a parse exception: \"a.\" read as [a],\n\"4 .1`\" as 4.1. ContentLanguageParser now expects the end of the value,\nand MimeVersionFieldImpl calls parseAll(). Comments and whitespace are\nskipped tokens, so a trailing comment stays accepted.\n"
    },
    {
      "commit": "c8ff400ceb7d28ee2cedb7c87fc190ff400bb9b3",
      "tree": "66bb588eeca8065234ae2f17a9b073457ed24ee8",
      "parents": [
        "f3672e214fe37e17dede8e6c1470d101abbeedc0"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 15:24:38 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 03 16:19:51 2026 +0300"
      },
      "message": "[FIX] StructuredFieldParser: a quoted pair escapes one character, also in a comment\n\nThe outermost comment had no quoted pair rule, so \"\\)\" ended it:\n\"http://example.com/(a\\) b)c\" was read as \"http://example.com/b)c\".\n\nA quoted pair also matched the backslash and the whole rest of the\nvalue, as ANY was one character or more: a quoted pair in a quoted\nstring or a nested comment swallowed the closing quote, and the value\nread as null.\n"
    },
    {
      "commit": "f3672e214fe37e17dede8e6c1470d101abbeedc0",
      "tree": "c46d1201cae03fa3da02238456db05ab233f98fc",
      "parents": [
        "dc1e6419125f06210892034ee70b3658f7ba7fab"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 15:18:46 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Tue Nov 03 16:19:51 2026 +0300"
      },
      "message": "Requalify closed fuzzing findings on master\n\nOne test per finding, with more inputs than the original regression\ntests: core and DOM descriptors on duplicate Content-Type and\nContent-Transfer-Encoding fields, case insensitive Content-Disposition\nparameters, encoded word lookalikes in subjects and display names,\nobsolete zones and one digit years in lenient dates, header list and\nindex agreement on non-ASCII case folding, encoded word lookalike\nboundaries, formatDate with offsets holding seconds, CRLF mbox files.\n"
    },
    {
      "commit": "dc1e6419125f06210892034ee70b3658f7ba7fab",
      "tree": "5ded6348927193ceec4c1470d7b5e6c55b89af57",
      "parents": [
        "b5eb7aa8f3bd6311ef6598a1ac4489c9b3968674"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 09:56:49 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[PATCH 3/3] fixup! [FIX] MimeUtil#fold: never write a whitespace only line\n"
    },
    {
      "commit": "b5eb7aa8f3bd6311ef6598a1ac4489c9b3968674",
      "tree": "060b173c6ad249a67db8a08dc75c6706c25923ab",
      "parents": [
        "6097c81bf6cc0978d29970eaebf4e2f4199a391d"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 09:56:49 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[PATCH 2/3] fixup! [FIX] DateTimeParser: reject years of more than 9 digits\n"
    },
    {
      "commit": "6097c81bf6cc0978d29970eaebf4e2f4199a391d",
      "tree": "9e47058c1dbee5b059d7478a208ed797d3110702",
      "parents": [
        "a586654f668cdf890ea0b6c9986389da50645d66"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 09:56:28 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "fixup! [PERF] DateTimeFieldLenientImpl: parse the common date form with a simple formatter\n"
    },
    {
      "commit": "a586654f668cdf890ea0b6c9986389da50645d66",
      "tree": "82c7297d779329228f9e7e1ac7d06249c9ba2fd0",
      "parents": [
        "5923886c1cdb1025f6da88856ac595d56e8bbf0a"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:29:50 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[PERF] DateTimeFieldLenientImpl: parse the common date form with a simple formatter\n\nNearly every mail dates itself as [Ddd, ]d Mon yyyy HH:MM:SS +HHMM.\nA cheap check of the shape now sends such a value to a formatter without\nthe optional sections and the lenient resolution of RFC_5322. Any other\nvalue, or an out of range one, goes to RFC_5322 as before.\n\nAn oracle test checks that whenever the simple form answers, RFC_5322\ngives the same instant, on every combination of boundary values and on\nmutated dates.\n"
    },
    {
      "commit": "5923886c1cdb1025f6da88856ac595d56e8bbf0a",
      "tree": "39df30e41254fac9c542e5863acab245f34543a5",
      "parents": [
        "d38c667be4068a9324cab2d0d8b9056c7548241d"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:09:18 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[TEST] maxHeaderCount: a header of exactly maxHeaderCount fields is accepted\n\nPins the documented semantic at the boundary, like the other per entity\nlimits: maxHeaderCount fields are accepted, one more is rejected.\n"
    },
    {
      "commit": "d38c667be4068a9324cab2d0d8b9056c7548241d",
      "tree": "17c76bb5937070ef6df8e6d8fb73f80a9c0f931b",
      "parents": [
        "ee64b73637900be6cba160dd8795633aec803ab6"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:03:24 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] MimeTokenStream: a zero-length read does not consume the multipart body raw\n\nA read(b, off, 0) on the content stream at T_START_MULTIPART marked it\nas used, so next() took the multipart as consumed raw and skipped its\npreamble and every body part, although nothing was read.\n\nA zero-length read now returns 0 without touching the stream state, as\nInputStream specifies. The getInputStream javadoc tells what reading at\nT_START_MULTIPART does.\n"
    },
    {
      "commit": "ee64b73637900be6cba160dd8795633aec803ab6",
      "tree": "1979f699f4e8789f34db6d4fd2791f543642b0c8",
      "parents": [
        "a5a7ee4a832c150377aa84be595c33da04df178d"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:02:50 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] maxHeaderLen: accept a header field of exactly maxHeaderLen bytes\n\nDefaultFieldBuilder#append rejected a field whose length reached the\nlimit, while MimeConfig documents rejecting a field exceeding it.\n"
    },
    {
      "commit": "a5a7ee4a832c150377aa84be595c33da04df178d",
      "tree": "09fd15ef8183a0926795d66b86ca200ccdedd786",
      "parents": [
        "6db025c179898b33621f5812ff35b9eade8a86c4"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:02:25 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] maxContentLen: report an exceeded limit as a MimeException\n\nAs documented, parsing a body longer than maxContentLen is terminated\nwith a MimeException, but a plain IOException was thrown: callers taking\nIOException for an I/O failure would retry a message to reject.\n\nLimitedInputStream now throws a MimeIOException wrapping the new\nMaxContentLengthLimitException, and next() unwraps it when draining a\ncontent.\n"
    },
    {
      "commit": "6db025c179898b33621f5812ff35b9eade8a86c4",
      "tree": "0e2cd1aea10dd42ec59e8b15301105f1ffc626af",
      "parents": [
        "20133b9c3fc3662c25e2bad026fe7d1e3e8b17b2"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:01:48 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] maxContentLen: apply the limit to the whole content, whatever the number of streams\n\nEach getInputStream() call, and the drain of next(), wrapped the content\nin a new LimitedInputStream counting from 0: a body read through two\nstreams, or partly read then skipped by next(), could exceed\nmaxContentLen without error.\n\nThe limited stream is now created once per content and shared by every\nreader of it.\n"
    },
    {
      "commit": "20133b9c3fc3662c25e2bad026fe7d1e3e8b17b2",
      "tree": "4622a1ba47fcdf6d211536e8a8972a58260dcd00",
      "parents": [
        "724ec010d4dcb1885c0a86fc0c8e27d008a4add2"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:01:24 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] maxContentLen: 0 disables the check, a body of exactly the limit is accepted\n\nAs documented, a non positive maxContentLen disables the check, but 0\nmade every body, even an empty one, fail with \"Input stream limit\nexceeded\".\n\nLimitedInputStream threw as soon as the limit was reached, before the\nread reporting the end of the stream: a body of exactly maxContentLen\nbytes was rejected. It now reads one more byte at the limit and throws\nonly when there is one.\n"
    },
    {
      "commit": "724ec010d4dcb1885c0a86fc0c8e27d008a4add2",
      "tree": "87752637986bc4202e5ef5a7227ccac4b7ef3ecd",
      "parents": [
        "8b466226e36784ab06d8e6ac190cca83a03dff4a"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:00:43 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] Mailbox#hashCode: fold the domain the way equals compares it\n\nequals compares domains with String#equalsIgnoreCase, which folds each\ncharacter to the lower case of its upper case, while hashCode hashed\ndomain.toLowerCase(Locale.US). Both disagree on some non-ASCII\ncharacters (\"İ\" or \"ı\" vs \"i\", \"ſ\" vs \"s\", \"ς\" vs \"σ\", \"µ\" vs \"μ\"):\nequal mailboxes got different hash codes, and a HashSet kept the same\nrecipient twice.\n\nLangUtils#hashCodeIgnoreCase hashes the same per code point folding.\n"
    },
    {
      "commit": "8b466226e36784ab06d8e6ac190cca83a03dff4a",
      "tree": "15eb3ec3b60dee34f38a908c2ff89e94131c4328",
      "parents": [
        "78ad364cad213b3110561ce9c599dcd41e90f1f1"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 17:00:02 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] EncoderUtil#encodeEncodedWord: choose the charset when the given one cannot encode\n\nA charset that only decodes (ISO-2022-CN, x-JISAutoDetect), for\ninstance taken from a received part, made encodeEncodedWord throw an\nundocumented UnsupportedOperationException. The charset is now chosen\nautomatically, as for a null charset.\n"
    },
    {
      "commit": "78ad364cad213b3110561ce9c599dcd41e90f1f1",
      "tree": "568b5bb48be942812f7aa6e0ef51d713dead8f8f",
      "parents": [
        "8e19b6c071e51e45a2708de05189929cf62e99a5"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 16:59:44 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] StorageBodyFactory#textBody: fall back to US-ASCII for a charset that cannot encode\n\nAs documented, a MIME charset whose Java charset cannot encode\n(ISO-2022-CN, x-JISAutoDetect) is replaced with US-ASCII, with a\nmonitor warning. The check had been lost, the body then failed at write\ntime with an UnsupportedOperationException.\n"
    },
    {
      "commit": "8e19b6c071e51e45a2708de05189929cf62e99a5",
      "tree": "be446da5529665474487adb020e71cb6b56b864a",
      "parents": [
        "bfda181df1d6f9db9ae5f6330ebb5721c6b08267"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 16:59:44 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] maxLineLen: accept a line of exactly maxLineLen bytes\n\nA line reaching the limit was rejected: with the default 1000, the\n998 characters + CRLF lines RFC 5322 allows failed to parse.\n\nA terminated line, its terminator counting as one byte, is now rejected\nonly when longer than maxLineLen; a line not terminated yet only when\nits bytes alone are.\n"
    },
    {
      "commit": "bfda181df1d6f9db9ae5f6330ebb5721c6b08267",
      "tree": "399dfa9875d81b4f95076eb7fa0ded19ac5ab10f",
      "parents": [
        "624de5d931c48063f46877c1ad0f37a58a6459cf"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 16:59:44 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] DOM builders: encode text in UTF-8 when the charset only decodes\n\nSome charsets the JDK knows only decode: ISO-2022-CN, x-JISAutoDetect.\nA received text part may carry one, and reusing its charset to build a\nnew text body (footer, reply, forward) failed at write time with an\nUnsupportedOperationException from Charset#newEncoder.\n\nCharsetUtil#forEncoding returns UTF-8 for such a charset. Every place\ncreating a text body from a String uses it, so that both the bytes and\nthe charset label are UTF-8: BasicBodyFactory text bodies,\nSingleBodyBuilder#buildText, BodyPartBuilder and Message.Builder\nsetBody(String, String, Charset), MultipartBuilder#addTextPart.\n\nBodies built from bytes keep their charset: decoding them is fine.\n"
    },
    {
      "commit": "624de5d931c48063f46877c1ad0f37a58a6459cf",
      "tree": "d6f320ce6226292d6b1dc378e3f1d2361103b08d",
      "parents": [
        "54f099672cb46ca879c812c7db7ae272f69909a7"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Sun Nov 01 16:54:20 2026 +0000"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[PERF] MimeBoundaryInputStream: look for the delimiter at line starts only\n\nThe delimiter was searched with Quick Search over the whole buffered\ndata, then rejected when not at a line start. For a boundary made of\none repeated character and a body of runs of that character shorter\nthan the delimiter, Quick Search shifts by one byte and restarts the\ncomparison at each alignment: O(body length x boundary length).\n\nThe delimiter is now only compared at the line starts, found with a\nbyte search for LF. The search position, and how much of the delimiter\nalready matched there, are kept across refills: every byte is compared\na bounded number of times whatever the boundary and the size of the\nreads from the source, so the search is linear in the part size.\n\nOnly what may still turn out to be a delimiter, with its preceding line\nbreak, is held back from the reader, instead of boundary length + 2\nbytes.\n\nThis also stops accepting a delimiter in the middle of a line when the\nreader happened to stop right before it.\n"
    },
    {
      "commit": "54f099672cb46ca879c812c7db7ae272f69909a7",
      "tree": "bc644a3ff9ae88bbb4ba588b6bce0628368b73e0",
      "parents": [
        "2e060b7d4bd8218d5e6db6e36c16370a389d144c"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 16:17:06 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[REFACTORING] EncoderUtil: check RFC 2231 parameter names without a regex\n"
    },
    {
      "commit": "2e060b7d4bd8218d5e6db6e36c16370a389d144c",
      "tree": "2cfede87a1f9fd251ffeee2cc7e2e9e3a1a509e3",
      "parents": [
        "73fd0cdd7c281233264f21943b84db884806cc75"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 11:41:20 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] EncoderUtil#encodeHeaderParameter: reject invalid parameter names\n\nParameter names were written verbatim: a name holding \"*\" was read back\nas an RFC 2231 section, and \";\" or \"\u003d\" added parameters. Reject names\nthat are not tokens, and \"*\" outside of the RFC 2231 forms (name*,\nname*N, name*N*) that callers copying parsed parameters may pass.\n"
    },
    {
      "commit": "73fd0cdd7c281233264f21943b84db884806cc75",
      "tree": "08913060ea1c436b14266e898e25c247f39f80ff",
      "parents": [
        "5e7760c94c3438e64a3020ef1738837d2ecb822a"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 11:38:10 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] DefaultMessageWriter: never encode a multipart body\n\nA multipart whose Content-Transfer-Encoding said base64 or\nquoted-printable was written encoded, while the parser never decodes a\nmultipart (RFC 2045 section 6.4): every body part was lost after one\nwrite/parse round trip. Write multipart bodies as is.\n"
    },
    {
      "commit": "5e7760c94c3438e64a3020ef1738837d2ecb822a",
      "tree": "a39cd3ec2782ea3bd833ca2433f049c3247726e0",
      "parents": [
        "dde230bdbf167a50861962938ad3545bb3147017"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 11:36:55 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] AddressFormatter: no \"@\" when the domain is neutralized to nothing\n\nencodeAddressDomain can drop every character of a malformed domain, and\nthe formatter then wrote \"local@\" with nothing after the \"@\". Treat\nsuch a mailbox as a domainless one.\n"
    },
    {
      "commit": "dde230bdbf167a50861962938ad3545bb3147017",
      "tree": "47d1a79ffe68d316260a0f1c630d638d83b6ef62",
      "parents": [
        "01cf28517ffca504f05b0879fec2ae7bcb47b2a0"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 11:35:49 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] RawField(name, body): reject names that are not RFC 5322 field names\n\nOnly the body was checked for line breaks: a name holding \":\", a\nwhitespace or a line break was written verbatim and changed the fields\nof the header, which allowed header injection through the name. The\nlenient constructor used by the parser is unchanged.\n"
    },
    {
      "commit": "01cf28517ffca504f05b0879fec2ae7bcb47b2a0",
      "tree": "1ffbf82b55aa88a2f5685672cd90b747d927cb55",
      "parents": [
        "040b76676aade44ad73fa956e71e928caaeeabec"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 09:43:20 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] AbstractEntity#getMimeType: ignore a parent multipart type without boundary\n\nThe digest default (message/rfc822) was applied below a part whose\nmultipart/digest type was discarded for lack of a boundary, so the\nembedded message of such a part reported message/rfc822 while it is\nparsed as text/plain.\n"
    },
    {
      "commit": "040b76676aade44ad73fa956e71e928caaeeabec",
      "tree": "79a72a0b51b122b8078e2f7a606970ff8925044c",
      "parents": [
        "fc969233968d30b978f7253e4f837c8f104549f9"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 09:38:27 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] MimeUtil#fold: never break right before a fold already in the value\n\nfold could break at a whitespace run that directly precedes an existing\nCRLF fold, writing a whitespace only line (\"Name:\" CRLF \" \" CRLF HTAB\n\"value\"). Do not break there, and do not count a lone CRLF as line text.\n"
    },
    {
      "commit": "fc969233968d30b978f7253e4f837c8f104549f9",
      "tree": "4732210689a1c0fa86c404525d2036a352df9cb6",
      "parents": [
        "c67be25c55907ad6f5e99e3bfc570802c5a71642"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 09:37:14 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] RawField#checkLineBreaks: reject a whitespace only continuation line\n\nOnly the first character after CRLF was checked, so a fold followed by\nwhitespace only (\"\\r\\n\\t\", \"x\\r\\n \\r\\n y\") was accepted and\nFields.* and DefaultMessageWriter wrote a whitespace only header line.\n"
    },
    {
      "commit": "c67be25c55907ad6f5e99e3bfc570802c5a71642",
      "tree": "ae6ff04823db3b33d7a58415ec7312281e2e7f29",
      "parents": [
        "76e52204bcaeb9e9968e056856b09f6979a11304"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 09:36:31 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] DefaultBodyDescriptorBuilder#newChild: hand children the resolved type\n\nnewChild passed the raw type of a part whose multipart Content-Type had\nbeen discarded for lack of a boundary, so the child of a multipart/digest\npart without boundary defaulted to message/rfc822 again and\nDefaultMessageBuilder nested a message level MimeTokenStream does not.\nPass the type build() resolves.\n"
    },
    {
      "commit": "76e52204bcaeb9e9968e056856b09f6979a11304",
      "tree": "2c15d8c31a6ac3d9ca3edee1cbae7c0184941658",
      "parents": [
        "9004b30cb4be7d520f4834bbcf2bc47f9cb38163"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 09:35:30 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] RawFieldParser#parseField: read the field name verbatim\n\nThe name was read as a structured token, with comment handling: the\ncomment of \"Subj(x)ect: x\" was dropped, so it was read as a Subject\nfield that RFC 5322 software sees as an unknown field, and an unbalanced\n\"(\" swallowed the colon so the field was lost. The name is now the text\nbefore the first colon, minus the whitespace around it.\n"
    },
    {
      "commit": "9004b30cb4be7d520f4834bbcf2bc47f9cb38163",
      "tree": "01c9e91103964dc846fc57c4867e398638945563",
      "parents": [
        "436b4ed470c768336e0d312ef341fcb30cdc3544"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 09:34:10 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] EncoderUtil#encodeAddressDomain: neutralize instead of rejecting\n\nRejecting a domain that is neither a dot-atom nor a domain literal made\nAddressFormatter and Fields.from/to/cc throw for mailboxes that\nLenientAddressParser read from malformed addresses (\"a@[\",\n\"a@example.com.\"), so received messages could not be written back.\nDrop only the characters that would change the structure of the header\n(\",\" adding recipients, quotes, comments, whitespace, controls).\n"
    },
    {
      "commit": "436b4ed470c768336e0d312ef341fcb30cdc3544",
      "tree": "e648ed22624c143ed5265c45e73d904e24d7b25e",
      "parents": [
        "f98fb2ffa044df36c2b4080cee6e56ba3a298c06"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 07:37:36 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] EncoderUtil: quote display names whose whitespace would not read back\n\nisAtomPhrase accepted leading, trailing, repeated whitespace and tabs,\nwhich readers drop or collapse into one space, so \"J  J\" or \" a\" read\nback as \"J J\" and \"a\". Only words separated by a single space are\nwritten as an atom phrase; other names are quoted.\n"
    },
    {
      "commit": "f98fb2ffa044df36c2b4080cee6e56ba3a298c06",
      "tree": "9b6450e910c14201cca67d186c2e09d549a60613",
      "parents": [
        "23794d504d1f7fa7e253d472155c797285f35a06"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 07:35:54 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] AddressFormatter: reject a domain that is not a dot-atom or a domain literal\n\nThe domain of a Mailbox was written verbatim, so a domain holding \",\"\n(\"example.com, other@example.org\") added recipients to the written\nheader. Check it with the new EncoderUtil#encodeAddressDomain.\n"
    },
    {
      "commit": "23794d504d1f7fa7e253d472155c797285f35a06",
      "tree": "63d9182e258217baf5752a0ae472f3e5c42c8858",
      "parents": [
        "745a979415696fa2ccd7d4ceea2014821bce0c89"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 07:34:15 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] DefaultAddressParser: RFC 2047 decode group names\n\nOnly mailbox display names were decoded, so a group name kept its\nencoded form, unlike with LenientAddressParser.\n"
    },
    {
      "commit": "745a979415696fa2ccd7d4ceea2014821bce0c89",
      "tree": "8ccac40d549dd374cbb411a5d320fc4be735af70",
      "parents": [
        "819128ec7d997d18427117dc001e73b9228d4fc9"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 05:57:50 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] EncoderUtil: drop CR and LF from a quoted local part\n\nThey can be neither qtext nor a quoted-pair, so AddressFormatter wrote\nthem raw inside the quoted string and the reader dropped them, changing\nthe address (\"\\\" CR read back as \"\"). Drop them when quoting.\n"
    },
    {
      "commit": "819128ec7d997d18427117dc001e73b9228d4fc9",
      "tree": "928db98def426e270e1fe3e3380ed334ea943424",
      "parents": [
        "41f5466a667f45fd4d1fd623d60e61cc9640df01"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 05:57:06 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] EncoderUtil: encode display names holding CR or LF instead of rejecting them\n\nRejecting them made AddressFormatter and Fields.from/to/cc throw for a\nparsed mailbox whose decoded display name holds a line break\n(\"\u003d?us-ascii?q?a\u003d0Ab?\u003d\"), so a received message could not be written\nback. Encode them as an encoded word: never written raw, read back\nunchanged.\n"
    },
    {
      "commit": "41f5466a667f45fd4d1fd623d60e61cc9640df01",
      "tree": "e326a9064a947b69e906d9329d40df40c7ae1d34",
      "parents": [
        "74fedaaccaf338b050f1c7f7e72487755b0241c3"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 05:08:57 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] Lenient Content-Type parsing: only trim whitespace\n\nString#trim also stripped control characters, so \"message/rfc822\\x05\"\nwas a message/rfc822 for the lenient parsers (ContentTypeFieldLenientImpl,\nFallbackBodyDescriptorBuilder) but not for the strict one, and the same\nmessage got two MIME structures. Trim CR, LF, SP and HT only.\n"
    },
    {
      "commit": "74fedaaccaf338b050f1c7f7e72487755b0241c3",
      "tree": "36834c4237e857832f018e4f925c9d483186e66e",
      "parents": [
        "6696f6e4dfd07100b2af45f8a126c771efa070d7"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 05:07:28 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] DefaultFieldBuilder: header length limit must not count CRLF as 2 bytes\n\nmaxHeaderLen counted the line terminators, so a field accepted with LF\nline ends was rejected once DefaultMessageWriter ended its lines with\nCRLF. Count each line as if it ended with a single LF, as the line\nlength limit does.\n"
    },
    {
      "commit": "6696f6e4dfd07100b2af45f8a126c771efa070d7",
      "tree": "e8a0d69bfa7882720d85ef0fab14f9c7b0ce10e4",
      "parents": [
        "2496eb9915c3d6f02e6a9cc6857efab79e3a0d97"
      ],
      "author": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Fri Oct 30 05:05:46 2026 +0100"
      },
      "committer": {
        "name": "Benoit TELLIER",
        "email": "btellier@linagora.com",
        "time": "Mon Nov 02 13:20:46 2026 +0300"
      },
      "message": "[FIX] MimeStreamParser: report the end of a headless header once\n\nWhen the first step of parseHeadless failed, the token stream stayed on\nT_END_HEADER, which MimeStreamParser reported again after its own\nendHeader: DefaultMessageBuilder threw IllegalStateException (\"Internal\nstack error\") instead of the MimeException. Call next() first, which\nthrows that exception.\n"
    }
  ],
  "next": "2496eb9915c3d6f02e6a9cc6857efab79e3a0d97"
}
